The meteorological structured data includes a variety of data types, such as ground hourly data, ground minute data, etc. Time attributes and spatial attributes need to be used in the analysis and call. With the increasing amount of data, queries and statistics under the conditions of high concurrency and complex statistics have strict requirements on the efficiency of the database. In this context, the traditional relational database is gradually difficult to meet the needs of real-time applications. This paper uses different types of distributed databases and relational databases, combined with meteorological business use scenarios, to explore the performance of databases under various types, and study the relationship between the performance differences of different scenarios and different database architecture frameworks to improve the real-time response capability of data services.